This project is qualified for participation in the "National Pioneer Cup on Intelligent Computing – Shandong University of Science and Technology Selection
Cu-Sn Machine Learning Interatomic Potential
This repository contains a machine learning interatomic potential (MLIP) model for the Cu-Sn alloy system, developed using the Deep Potential Generator (DP-GEN) and DeepMD-kit frameworks. The trained model is provided as frozen_model.pb.
Using this DP model, we performed molecular dynamics (MD) simulations to compute the energy–volume (E–V) curve, elastic moduli, and phonon spectra of Cu-Sn compounds. The simulation results show excellent agreement with density functional theory (DFT) calculations, demonstrating that the developed model achieves DFT-level accuracy while maintaining significantly higher computational efficiency.
Software Used
DP-GEN: Automated active learning workflow for training interatomic potentials
DeepMD-kit: Neural network potential training package
LAMMPS: Molecular dynamics engine for simulation
Repository Structure
├── result # MD results: energy–volume curves, phonon spectra
│ ├── E-V.jpg # Energy-Volume curve
│ └── Phonon dispersion relation.jpg # Phonon dispersion relation
├── frozen_model.pb # Final trained Deep Potential model
└── Research on Efficient Molecular Dynamics Simulation Methods Based on Machine Learning Potentials/ # Experimental report and summary